No one has ever described deploying and managing Kubernetes to support app modernization initiatives as easy.
Getting Kubernetes up and running is one thing. Performing ongoing security and maintenance operations, and ensuring compute capacity is cost-optimized as the various autoscaling technologies baked into Kubernetes wrestle with each other is another.
Becoming proficient at these technologies often means large upfront investments in cloud infrastructure, laundry lists of certifications, and lengthy professional services engagements. Cost optimization often comes last. In many cases the costs and complexity of embarking on an app modernization initiative are just too daunting for many organizations to even begin.
With Amazon EKS Auto Mode, you can automate cluster management without deep Kubernetes expertise. EKS Auto Mode takes care of:
- Selecting optimal compute instances
- Dynamically scaling infrastructure resources
- Continuously optimizing costs
- Managing core add-ons, patches operating systems
- Integrating with AWS security services out of the box
If managing the operational burden of these areas wasn’t enough to keep engineering leaders up at night, organizations are also caught in a tug-of-war between rising cloud costs and ensuring application reliability. This leaves business stakeholders, software developers and platform engineers often at odds, leading to stalemates.
Achieving operational efficiency in Kubernetes requires new capabilities, built on analytics and automation, which can help ease the burden on people, and achieve both cost efficiency and application reliability.
Intelligent Cluster Autoscaling Reduces Infrastructure Costs #
EKS Auto Mode is a new feature that fully automates security, maintenance, and compute management for Kubernetes clusters. By offloading the operational overhead of managing the cluster infrastructure required to run production-grade Kubernetes applications at scale on AWS, EKS Auto Mode improves operational efficiency and helps optimize compute costs.
The compute capacity management feature of EKS Auto Mode analyzes the resource request values of unschedulable pods, determines the least-cost compute instances to meet application requirements, and quickly provisions the optimal instance type to run them.
Over time, those instances can become underutilized as some workloads scale down or are removed from the cluster. EKS Auto Mode looks for opportunities to reschedule these workloads onto a set of more cost-efficient EC2 instances, whether they are already in the cluster or need to be provisioned.
For those familiar with Karpenter and its reputation as a high-performance Kubernetes cluster autoscaler that was introduced by AWS in 2021 and donated to the Cloud Native Computing Foundation (CNCF), it is exactly what’s running under the hood of EKS Auto Mode. This underscores how tried and true this offering is as a compelling solution for those organizations looking to reduce their operational overhead.
Improper Pod Requests Contribute a Majority of the Waste #
Amazon EKS Auto Mode drastically simplifies the configuration of cluster autoscaling and can improve application availability and cluster efficiency to reduce cloud costs. However, it is not designed to address the more granular need for pod rightsizing, nor does it provide the ability to add and remove pods in response to bursty workloads.
According to a survey by the CNCF, 49% say Kubernetes has driven cloud spend up, and 70% say workload overprovisioning is the source of overspending. This improper workload resource management results in wasted resources that directly correspond to excessive cloud costs while leaving application performance and availability risks unaddressed.
Solving for these suboptimal pod requests can be challenging as developers and platform engineers spar over what these values should be, how often they should be updated, and ultimately, whose responsibility it is to perform what quickly becomes an excessive amount of manual toil. Kubernetes native solutions like the VPA are insufficient for solving these problems accurately at scale.
Autonomously Manage Resources With Machine Learning #
StormForge Optimize Live rightsizes pods automatically, using machine learning to analyze usage data from workloads to recommend the most efficient CPU and memory settings without risking performance or reliability. Recommendations can be automatically implemented to keep pods continuously rightsized as usage fluctuates.
In addition, StormForge works with the HPA to scale bi-dimensionally, recommending the optimal target utilization to ensure the HPA’s scaling behavior is consistent as the vertical size of the pods change. Because StormForge is managing the configuration of both the vertical scaling and the HPA target utilization, vertical and horizontal pod autoscaling work synergistically — eliminating thrashing and maximizing pod-level efficiency.
How EKS Auto Mode and StormForge are Better Together #
Amazon EKS Auto Mode is a great solution on its own, but it doesn't keep pods rightsized or scale them in the most efficient way. Using Amazon EKS Auto Mode alongside StormForge ensures efficient bin packing by streamlining the placement of pods on optimally sized nodes, which results in additional cloud cost savings.
By deploying Amazon EKS Auto Mode and StormForge together, you can maximize efficiency — automatically and intelligently.
Key Results:
- Reduce Kubernetes-related cloud costs and resource usage by 50% or more
- Ensure application performance and reliability for higher productivity and a positive customer experience
- Free up operational bandwidth from Kubernetes management to focus on innovation and other strategic activities
Learn more about maximizing efficiency on Amazon EKS with StormForge, or if you're ready to get started, Sign Up to try StormForge free for 90 days exclusively on AWS Marketplace.
